[Retracted] Financial Fraud Detection in Healthcare Using Machine Learning and Deep Learning Techniques

Table1

Limitations of machine learning techniques.

Model	Strength	Limitations
Bayesian	Provide better results in problems of binary classification and suitable for analyzing the real-time data	Required better detection related to the abnormal and expected behavior of fraud cases
Neural Network	Suitable for problems related to binary classification, mostly used for detecting the fraud	Required huge computation, can be denied for real-time operation, and retraining is essential in terms of newly arrived fraud cases
Decision Tree	Implementation is more straightforward with low power of computation and suitable for analyzing the real-time data	Overfitting may rise if the information of the underlying domain does not set in training data
Logistic Regression	Implementation is easy and fraud detection is based on historical data	Performance of classification is lacking when compared with methods of data mining
Linear Regression	When dependent and independent variables have an almost linear relationship, it generates an optimal result	Sensitive for the outliers and numeric value limitation
Support Vector Machine	The nonlinear problem of classification is solved with low power of computation and suitable for analyzing real-time data	Input data transformation results in difficulties while processing the data